Artificial Intelligence (AI) is as inescapable as the Atlanta pollen, but users are glossing over the extreme environmental and ethical issues as companies replace jobs and students trade hard work with AI-generated garbage. Generative AI is an unregulated beast that will take everything from entry-level jobs to authentic artwork and leave us with nothing if we do not change. AI squanders the resources it requires and must be trained on the backs and minds of actual people without their consent, often failing to provide helpful or accurate answers. Using generative AI platforms in place of a normal search engine is not only bad for the user due to potential misinformation, but is also unethical and lazy.
The training and maintenance of generative AI models utilize gargantuan amounts of energy and water and create ridiculously high carbon emissions. Quantifying exact numbers is difficult for the industry as a whole, but researchers estimate that in the next two years, AI could make up 0.5% of the world’s electricity consumption. This might not sound like much until you consider that this would mean AI would utilize as much as the entire country of Argentina.
A study from Cornell estimated that within the same timeframe, the amount of water utilized to cool AI data centers will reach half of England’s annual water consumption, or more than four times the usage of the entire country of Denmark. This is simply a waste of valuable resources that we cannot afford in the face of a growing climate crisis.
While the exact impact of the entire AI industry is difficult to measure, the impact of individual models is easier to track. For example, the training of GPT-3 created about 500 tons of carbon dioxide. This is equivalent to the average annual electricity consumption of over 127 American households and 108 cars. Recall that this is only for the training of a single version of a single AI model, and it dwarfs the impact that a single person has throughout their entire lifetime of living or driving.
The training of GPT-3 also used over 180,000 gallons of water, equivalent to the daily drinking water requirements of about 218,000 people. In a world where 1 in every 11 people lacks access to enough clean drinking water, this is both unethical and disturbing.
Beyond the environmental impact, the ethical concerns behind every single AI model are wide- reaching. In particular, the data that developers use to train these models is sourced broadly and without consent. In 2023, the Atlantic reported that Meta trained its Large Language Model (LLM) on thousands of stolen and pirated works, and experts are warning that developers are training AI on any content you have posted online that isn’t behind a login wall (such as a public Instagram account or a LinkedIn profile). Artists and writers across the board have made their anger public as massive companies use their works to train various models that generate both images and text, leading to over 30 copyright infringement cases in the United States alone.
Training these models on personal data, articles, stories, artworks and any other creations that users like you and me have shared online is inherently wrong, especially for posts made before the advent of generative artificial intelligence entering the mainstream. Those who shared their creations had no way to agree to companies training AI models on their work and retroactively opting out is hardly an option.
The information generative AI programs peddle is false far more often than many people realize. Researchers have found that GPT-3 agrees with false statements up to over a quarter of the time and cites incorrect sources up to 60% of the time. When the energy required for a single request from ChatGPT is ten times that of a single Google search and gives you the wrong answer, I question what scenario generative AI is ever better than a traditional search engine. Even if the massive amounts of water and energy AI demands or the art it steals doesn’t matter to you, the fact that you are quite likely to receive an incorrect answer should.
Do you really want to support technology that is stealing your data, destroying the environment and stunting your capabilities every time you use it?
If so, it is time to do some self-reflection.
If not, it is time to stop using AI.